Extract claims from a source document and verify them against a target to generate a drift report.
npx skills add https://github.com/davekim917/bootstrap --skill team-drift
Please help me install this Agent Skill. Skill: team-drift Source: https://github.com/davekim917/bootstrap/tree/main/plugins/workflow/skills/team-drift Command: npx skills add https://github.com/davekim917/bootstrap --skill team-drift
This Skill helps teams ensure two documents reflect each other by extracting all claims from a source-of-truth and verifying them against a target using two independent agents, then classifying mismatches by severity.
Provide the two documents—the source-of-truth and the target—and run the drift check to generate a drift report.
High-intent search queries and answers about installing and using this skill.
Document drift detection works by extracting all claims from a source-of-truth and verifying each against a target using two independent agents. Mismatches are classified as CONFIRMED, PARTIAL, DIVERGED, or MISSING and compiled into a drift report.
The best way to verify claims for alignment is using a multi-agent approach where two independent agents check the source-of-truth against the target. This dual-agent verification surfaces discrepancies and classifies drift severity.
You need to run a drift check after planning or building stages, during design reviews, or at any workflow stage where document alignment is required. It ensures the target document accurately reflects the source-of-truth.
To generate a drift report, provide the source-of-truth and target documents as context. The process extracts and verifies claims, then classifies mismatches as CONFIRMED, PARTIAL, DIVERGED, or MISSING using the drift-report-template.md format.
Yes, automated drift detection can be used without external dependencies. You provide the two documents as sole context, and the system handles claim extraction, verification via two independent agents, and drift report generation independently.
The drift classification categories indicate severity of mismatches between documents. CONFIRMED, PARTIAL, DIVERGED, and MISSING describe whether claims in the source-of-truth are fully matched, partially reflected, diverged, or absent in the target document.
pluginagentmarketplaceDetect data and model drift in production ML deployments.
ConsiliencyDetect drift between database schemas and ORM models in Python and TypeScript projects.
jsagirTrack grant opportunities across discovery, research, application, and submission stages.
lucasFR2Execute verification commands and compare output before completion claims.
